Smart CAR-T Nanosymbionts: archetypes and proto-models

Abstract

Personalized medicine has redefined cancer treatment by aligning therapies with each patient's unique biological profile. A key example is chimeric antigen receptor T-cell (CAR-T) therapy, in which a patient's own T cells are genetically modified to recognize and destroy cancer cells. This approach has delivered remarkable results in hematologic malignancies and is beginning to show promise in solid tumors and autoimmune diseases. However, its broader adoption is limited by major challenges, including complex manufacturing, high costs, limited efficacy in solid tumors, and potentially severe toxicities. Nanotechnology offers exciting possibilities to overcome many of these barriers. Engineered nanoparticles can improve gene delivery, target tumors more precisely, enhance immune cell function, and enable in vivo CAR-T production, reducing the need for labor-intensive ex vivo processes. However, despite this promise, translation into clinical settings remains difficult due to regulatory hurdles, scalability issues, and inconsistent reproducibility in human models. At the same time, artificial intelligence (AI), with its powerful algorithms for data analysis and predictive modeling, is transforming how we design, evaluate, and monitor advanced therapies, including the optimization of manufacturing processes. In the context of CAR-T, AI holds strong potential for better patient stratification, improved prediction of treatment response and toxicity, and faster, more precise design of CAR constructs and delivery systems. Leveraging these three technological pillars, this review introduces the concept of Smart CART Nanosymbionts, an integrated framework in which AI guides the design and deployment of nanotechnology-enhanced CAR-T therapies. We explore how this convergence enables optimization of lipid nanoparticle formulations for mRNA transfection, specific targeting and modification of the tumor microenvironment, real-time monitoring of CAR-T cell behavior and toxicity, and improved in vivo CAR-T generation and overcoming barriers in solid tumors. Finally, it's important we also address the ethical and regulatory considerations surrounding this emerging interface of living therapies and computational driven systems. The Smart CART Nanosymbionts framework (Figure 1:) represents a transformative step forward, promising to advance personalized cancer treatment toward greater precision, accessibility, and overall effectiveness.

Keywords: CAR-T therapy; artificial intelligence; deep learning; immunotherapy; machine learning; manufacturing; nanotechnology; personalized medicine.

Figure 1

Compares the use of ML and DL in optimizing the production of advanced therapies, such as CAR-T cells and nanoparticles. In the ML section the algorithm can be trained for feature extraction, incorporating clinical data (such as biomarkers and immune responses) and manufacturing data (such as ethanol concentration and total flow rate). These inputs are analyzed to enable early prediction of CRS and ICANS toxicity and to improve the efficiency and quality of nanoparticle production. The DL section presents a more advanced approach using, ANNs, CNNs and GNNs. These technologies analyze proteins, RNA sequences, cellular imaging, and spatial representations. Key applications include the design of nanoparticles with enhanced tumor penetration, the optimization of CAR-T cells with greater cytotoxicity and persistence(CAR-Toner and motifs analysis), and the classification of cells based on their sensitivity or resistance to CAR-T treatment. Additionally, image-based predictions help assess cell sorting (COSMOS), therapeutic response and immune synapse quality. CRS, Cytokine Release Syndrome; ICANS, Immune Effector Cell-Associated Neurotoxicity Syndrome; ANNs, Artificial Neural Networks; GNNs, Graph Neural Networks.

Figure 2

The conventional CAR-T cell manufacturing model involves a complex ex vivo process with multiple steps, including leukapheresis, T-cell selection, genetic modification using viral vectors, expansion, and cryopreservation before infusion, leading to high costs, long production times, and logistical challenges. The “Addition by Subtraction Model: Smart CAR-T Nanosymbionts “ model integrates AI and nanotechnology to streamline production by reducing process steps, replacing viral vectors with non-viral alternatives (e.g., nanoparticles), and leveraging in vivo genetic modification to enhance efficiency. AI-driven patient selection analyzes clinical, biochemical, and imaging data to predict response, while AGILE-based discovery of nanoparticles optimizes transfection and biodistribution for improved CAR-T functionality. AI-enhanced bioreactor control using nanosensors ensures real-time monitoring and quality assessment and technologies like COSMOS helps in label-free sorting cells, refining the final CAR-T product. Additionally, AI-driven protein analysis optimizes CAR structure by improving peptide-CAR interactions, refining co-stimulatory domains, and identifying new neoantigens to enhance efficacy. Post-infusion, AI assists in predicting and managing adverse effects, while nanoadjuvants dynamically regulate CAR-T function, mitigating toxicity, preventing exhaustion, promoting epitope spreading, and strengthening the immunological synapse (IS). This AI- and nanotechnology-driven approach enhances CAR-T therapy by improving safety, reducing costs, and increasing accessibility, marking a significant advancement in cancer immunotherapy. Created with BioRender.com.

Figure 3

Schematic representation of lipid, polymer, inorganic, and hybrid hydrophobic polymeric nanoparticles (NPs) and the possible advantages of using nanotechnology in CAR T cells (1): Tumor microenvironment remodeling: using indocyanine green nanoparticles plus infrared light irradiation to disrupt the ECM before CAR administration, using targeted nanocarriers with in vitro transcribed mRNA to reprogram TAMs and downregulate PD-L1, and using nanozymes and nanoparticle backpacks. (2) Improving T cell proliferation and lifespan with mesoporous silica micro-rods secreting IL-2, APC cell-membrane mimics, using RNA-LPX to activate T cells, and NPs linking APCs to prime and activate T cells. (3) Improving follow-up and resistance with genetic programming using mRNA nanocarriers for targeted gene expression and NBiTE generation and radiolabeled NPS to track T cells in vivo. APC, antigen-presenting cell; ECM, extracellular matrix; IL-2, interleukin-2; NBiTEs, nano-bispecific T cell engagers; NPs, nanoparticles; PD-L1, programmed cell death ligand-1; TAMs, tumor-associated macrophages. Created with BioRender.com. Taken with permission from: Baena JC, Pérez LM, Toro-Pedroza A, Kitawaki T, Loukanov A. CAR T Cell Nanosymbionts: Revealing the Boundless Potential of a New Dyad. Int J Mol Sci. 2024 Dec 7;25(23):13157. doi: 10.3390/ijms252313157. PMID: 39684867; PMCID: PMC11642191.

Figure 4

Illustrative schematic of the In Vivo Smart CAR-T Manufacturing, highlighting four key phases in targeted cancer therapy: 1. Infusion: Administration of smart lipid nanoparticles optimized with artificial intelligence models such as XGBoost for improved scalability, AGILE for enhanced transfection efficiency, and Nano-AI-QSAR to optimize delivery and gene expression in target organs. 2. In vivo manufacturing: Direct conversion of T cells into CAR-T cells within the patient, eliminating the need for ex vivo manipulation, reducing production time and costs. 3. CAR-T enhancement: Implementation of AI models to improve CAR-T cell persistence and cytotoxicity through strategies like CAR-Toner to prevent exhaustion, and AI-improved IS + costimulatory domain to enhance cytotoxicity and memory formation. 3. Smart nano-adjuvants: Finally, independent strategies for solid tumors can further be improve with AI(like target delivery) to act in key factors such as; Modulation of the tumor microenvironment using TGF-β inhibitors and IL-15 nanogels, target monitoring and toxicity control with magnetic nanoparticles and Dasatinib, and enhanced CAR-T response and persistence with PD-L1 inhibitors. Created with: Biorender.com. Abbreviations:.

Figure 5

Graphical Abstract of Addition by Subtraction: Smart CART Nanosymbionts: AI serves as the central integrative engine, leveraging machine learning and deep learning algorithms to drive decision-making, pattern recognition, and predictive modeling. This AI-driven framework enhances the design, function and specificity of nanoparticles—such as lipid and -based nanoparticles—and improves CAR-T cell therapy by optimizing patient selection, target specificity, response prediction, and toxicity control. Together, these elements enable the development of intelligent, adaptive, and personalized cancer treatments.